The project Common Voice of Mozilla, whose objective is to create a word database accessible to all, recently launched a call to move forward on certain languages representative of minorities, but also on French.
Launched in July 2017, Mozilla’s Common Voice project aims to make a voice database accessible to everyone to help develop services and applications.
Proposing, within interfaces, languages adapted to users is an important challenge, especially since voice recognition has become more widespread. However, there are still many gaps, and Mozilla needs all the help it can get to achieve its goal of providing a free and open database.
French among the least complete languages
To create a voice database, there is no secret, you need votes. With its Common Voice project, Mozilla wants to allow each user, regardless of the language they use, to be able to operate voice recognition. All languages are concerned, including dialects representative of minorities. But the task is difficult, and many languages are unfortunately under-represented in many existing databases, as Mozilla indicates in a recent press release.
“ The databases that power voice recognition services are owned by a few large companies, and some voices are largely underrepresented. Voices like those of less common accents, local minorities, but also people with disabilities. […] And for some languages, it’s even a desert: the data is non-existent »
Mozilla wants to remedy this problem with its own free and open source database. The needs are numerous, whether for various English accents or many languages including Indonesian, Hindi, Aragonese, Tajik, Sicilian and French.
Add your stone to the building with your voice
In order to carry out their work, Mozilla developers need a large quantity of samples. French being in great demand at the moment, you can participate in the collection by lending your voice as long as you are French-speaking. Whether you live in Belgium, Luxembourg, Switzerland, North America or North Africa, you can go to the project page and follow the guide!
To put it simply, we offer you a whole bunch of sentences to read out loud and intelligibly in order to create lots of little tracks which will be recovered and processed. Mozilla also asks to provide a certain amount of demographic data to improve the quality of the database and the accuracy of speech recognition, but, rest assured, all data is anonymized.
To make everything more fun, Mozilla allows you to see the current progress of all languages, consult numerous statistical data, or even compare your progress to that of other contributors.
Source : Goodtech
2